l e t t e r s

Meta-analysis of genome-wide association studies identifies eight new loci for in east Asians

Yoon Shin Cho1,46, Chien-Hsiun Chen2,3,46, Cheng Hu4,46, Jirong Long5,46, Rick Twee Hee Ong6,46, Xueling Sim7,46, Fumihiko Takeuchi8,46, Ying Wu9,46, Min Jin Go1,46, Toshimasa Yamauchi10,46, Yi-Cheng Chang11,46, Soo Heon Kwak12,46, Ronald C W Ma13,46, Ken Yamamoto14,46, Linda S Adair15, Tin Aung16,17, Qiuyin Cai5, Li-Ching Chang2, Yuan-Tsong Chen2, Yutang Gao18, Frank B Hu19, Hyung-Lae Kim1,20, Sangsoo Kim21, Young Jin Kim1, Jeannette Jen-Mai Lee22, Nanette R Lee23, Yun Li9,24, Jian Jun Liu25, Wei Lu26, Jiro Nakamura27, Eitaro Nakashima27,28, Daniel Peng-Keat Ng22, Wan Ting Tay16, Fuu-Jen Tsai3, Tien Yin Wong16,17,29, Mitsuhiro Yokota30, Wei Zheng5, Rong Zhang4, Congrong Wang4, Wing Yee So13, Keizo Ohnaka31, Hiroshi Ikegami32, Kazuo Hara10, Young Min Cho12, Nam H Cho33, Tien-Jyun Chang11, Yuqian Bao4, Åsa K Hedman34, Andrew P Morris34, Mark I McCarthy34,35, DIAGRAM Consortium36, MuTHER Consortium36, Ryoichi Takayanagi37,47, Kyong Soo Park12,38,47, Weiping Jia4,47, Lee-Ming Chuang11,39,47, Juliana C N Chan13,47, Shiro Maeda39,47, Takashi Kadowaki10,47, Jong-Young Lee1,47, Jer-Yuarn Wu2,3,47, Yik Ying Teo6,7,22,25,41,47, E Shyong Tai22,42,43,47, Xiao Ou Shu5,47, Karen L Mohlke9,47, Norihiro Kato8,47, Bok-Ghee Han1,47 & Mark Seielstad25,44,45,47

We conducted a three-stage genetic study to identify have been identified for T2D10,11. Despite these advances toward a susceptibility loci for type 2 diabetes (T2D) in east Asian better understanding of the genetic basis of T2D, its heritability has populations. We followed our stage 1 meta-analysis of eight not been fully explained12. In addition, most of the T2D loci were T2D genome-wide association studies (6,952 cases with T2D detected initially in population samples of European origin, with the

Nature America, Inc. All rights reserved. All rights Inc. America, 2 Nature and 11,865 controls) with a stage 2 in silico replication exceptions of KCNQ1, UBE2E2 and C2CD4A-C2CD4B, which were analysis (5,843 cases and 4,574 controls) and a stage 3 de novo first identified in studies of east Asian populations13–15. Additional

© 201 replication analysis (12,284 cases and 13,172 controls). efforts involving east Asian populations identified variants associated The combined analysis identified eight new T2D loci reaching with T2D at the SPRY2, PTPRD and SRR loci5,16,17. However, these genome-wide significance, which mapped in or near GLIS3, associations need more validation from additional studies of east PEPD, FITM2-R3HDML-HNF4A, KCNK16, MAEA, GCC1-PAX4, Asians as well as in studies of other populations. A large meta-analysis PSMD6 and ZFAND3. GLIS3, which is involved in pancreatic in east Asians would be expected to identify new genetic associations beta cell development and insulin expression1,2, is known and provide insights into T2D pathogenesis. In addition to differences for its association with fasting glucose levels3,4. The evidence of in the allele frequencies between east Asians and Europeans, which an association with T2D for PEPD5 and HNF4A6,7 has been may affect the power to detect associations in these populations, T2D shown in previous studies. KCNK16 may regulate glucose- epidemiology also differs considerably between European populations dependent insulin secretion in the pancreas. These findings, and east Asian populations. In east Asians, the rates of diabetes are derived from an east Asian population, provide new often higher at lower average body mass indices (BMIs)18, suggesting perspectives on the etiology of T2D. that some different pathways may be involved in pathogenesis of T2D in east Asians and Europeans. T2D is a major public health problem with a rapidly rising global To discover new T2D loci, we conducted a three-stage association prevalence8. The development of T2D is influenced by diverse factors, study in individuals of east Asian descent (Supplementary Fig. 1). and decades of epidemiological studies have linked obesity, hyper- We performed the stage 1 meta-analysis by combining eight T2D tension and dyslipidemia with the risk of T2D9. It is also known that genome-wide association studies (GWAS) participating in the Asian T2D has considerable heritability. Within only the last 3 years, genetic Genetic Epidemiology Network (AGEN) consortium (6,952 cases studies have produced a rapidly lengthening list of loci harboring and 11,865 controls) with association data for 2,626,356 imputed and disease-predisposing variations10. To date, genetic variants at 45 loci genotyped autosomal SNPs, and we used the inverse-variance method

A full list of affilations are at the end of article. Received 12 April; accepted 2 November; published online 11 December 2011; doi:10.1038/ng.1019

Nature Genetics VOLUME 44 | NUMBER 1 | JANUARY 2012 67 l e t t e r s

for fixed effects for the statistical analyses (Supplementary Table 1). near FITM2-R3HDML-HNF4A identified in our study (rs6017317) is All imputed and genotyped SNPs (minor allele frequency (MAF) > not in strong LD with the HNF4A P2 promoter SNPs (r2 = 0.23–0.25, 0.01) passed quality control filters in each of the eight stage 1 datasets D′ = 0.50–0.54 between rs6017317 and rs1884613 or rs2144908 based prior to conducting the meta-analysis (Supplementary Table 2). The on phase 2 CHB/JPT HapMap data), indicating that rs6017317 is a genomic control inflation factor (λ) for the meta-analysis was 1.046 new T2D signal in the 20q13.12 region where HNF4A resides. (and was less than 1.062 for each of the individual studies), indicat- The other five loci reaching genome-wide significance in our ing that the results seen in stage 1 were probably not the result of study have not previously been reported in the context of any meta- population stratification (Supplementary Fig. 2). Individuals from bolic traits, including the loci mapped in or near KCNK16, MAEA, each component study that participated in stage 1 mainly clustered GCC1-PAX4, PSMD6 and ZFAND3. KCNK16, which is expressed together with the samples from the CHB/JPT HapMap population predominantly in the pancreas, encodes a potassium channel pro- in the principal component analysis plot (Supplementary Fig. 3), tein containing two pore-forming P domains23. In pancreatic β cells, further showing the similarity in ethnicity between the stage 1 samples. potassium channels that are inhibited by ATP regulate glucose- Most signals showing strong evidence for T2D associations were dependent insulin secretion. Among the variants in strong LD with in known T2D (Fig. 1). Stage 1 P values, odds ratios (ORs) the signal reaching genome-wide significance in KCNK16 (rs1535500) and average risk allele frequencies for 45 previously reported T2D- is rs11756091 (r2 = 0.977, D′ = 1.0 based on phase 2 CHB/JPT HapMap associated SNPs are listed in Supplementary Table 3. data), which encodes a substitution of proline to histidine in two After removing known T2D variants, we selected 297 SNPs from isoforms of KCNK16. This variant or others influencing KCNK16 independent loci from the stage 1 meta-analysis based on our arbi- may result in the defective regulation of potassium channel activity trary inclusion criteria for in silico follow-up replication: meta-analysis that contributes to the etiology of T2D24. MAEA encodes a P < 5 × 10−4 (based on the divergence between the observed and that has a role in erythroblast enucleation and in the development expected P values on the quantile-quantile plot; Supplementary of mature macrophages25. A gene-set analysis of the stage 1 P values Fig. 2), heterogeneity P > 0.01 and at least seven studies having been using GSA-SNP26 indicated that MAEA belongs to a group of genes included in the meta-analysis (Supplementary Table 4). We took a that previously showed significant association with T2D and includes total of 3,756 SNPs, including the 297 selected SNPs and their proxies IDE, which is located at a known T2D susceptibility locus27 (stage 1 (r2 > 0.8 based on phase 2 CHB/JPT HapMap data), forward to stage 2 P = 1.41 × 10−7 for rs6583826 at the IDE locus in this study). The (in silico replication) in three independent GWAS (5,843 cases and GRIP-domain–containing protein that is encoded by GCC1 might 4,574 controls). After a meta-analysis that combined stage 1 and have a role in the organization of the trans-Golgi network, which is 2 data for 3,756 SNPs, we selected the 19 SNPs that showed the involved in membrane transport28. PAX4, which is only 30 kb away most compelling evidence for association (stage 1 and 2 combined from GCC1, is a good candidate for T2D given its involvement in P < 10−5) (Supplementary Table 5) for stage 3 de novo genotyping in pancreatic islet development. PAX4 was recently implicated in a up to 12,284 cases and 13,172 controls recruited from five independ- Japanese individual with maturity onset diabetes of the young29. The ent studies (Supplementary Tables 1 and 2). This resulted in eight expression product of PSMD6, which acts as a regulatory subunit new T2D loci that reached genome-wide significance in the combined of the 26S proteasome, is probably involved in the ATP-dependent meta-analysis across all three stages (Table 1 and Fig. 2). degradation of ubiquitinated proteins30. Although the function of Three of these T2D-associated loci were previously associated with ZFAND3 has not been fully elucidated, it is noteworthy that a member metabolic traits or related diseases or were suggestively associated of the same gene family, ZFAND6, is present along with FAH at a Nature America, Inc. All rights reserved. All rights Inc. America, 2 Nature with T2D. We detected one such locus within an intron of GLIS3, previously detected T2D locus31. We examined whether eight new loci a gene that is highly expressed in islet beta cells. The coding prod- are potentially associated with T2D through an effect on obesity, as © 201 uct of this gene, a Krüppel-like zinc finger transcription factor, has is the case with FTO32. All of the T2D association signals we initially been proposed as a key player in the regulation of pancreatic beta cell detected remained after adjustment for BMI (Supplementary Table 6), development and insulin gene expression1,2. SNPs in high linkage indicating that the associations with T2D of these eight loci are not disequilibrium (LD) with this locus have been implicated in asso- mediated through an effect on obesity. ciation with type 1 diabetes (T1D)19 and fasting plasma glucose3. In addition to the eight loci reaching genome-wide significance, we The second such locus, on 19q13, is located in an intron of PEPD. identified two loci showing moderate evidence (combined P < 10−6) Several SNPs (lead SNP: rs10425678) in this gene were previously of association with T2D, including WWOX and CMIP loci (Table 1). associated with T2D in a Japanese population5. However, the SNP We obtained the association results for these ten loci in GWAS data in PEPD identified in our study (rs3786897) is not in LD with those from up to 47,117 European samples generated by the DIAGRAM identified in the Japanese population (r2 = 0.008, D′ = 0.143 between consortium (DIAGRAM+ is the current version of dataset)31. rs3786897 and rs10425678 based on phase 16 2 CHB/JPT HapMap data), and our GWAS CDKAL1 CDKN2A/2B data do not support an association for T2D KCNQ1 P HHEX/IDE

with rs10425678 (P = 0.528). The third such 10 8 IGF2BP2 SLC30A8 KCNJ11 FTO

signal is near FITM2-R3HDML-HNF4A. –log FITM2 may be involved in accu- mulation20, and the function of R3HDML 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 171819 20 2122 is not known. Mutations in HNF4A cause maturity onset diabetes of the young type 1 Figure 1 Genome-wide Manhattan plot for the east Asian T2D stage 1 meta-analysis. Shown are (ref. 21). Common variants in the P2 pro- the –log10 P values using the trend test for SNPs distributed across the entire autosomal genome. moter region of this gene (rs1884613 and The red dots at each locus indicate the signals with P < 10−6 detected in the genome-wide meta- rs2144908) have been associated with T2D analysis. A total of 1,934,619 SNPs that were present in at least five stage 1 studies were used to in a population-specific manner6,22. The SNP generate the plot.

68 VOLUME 44 | NUMBER 1 | JANUARY 2012 Nature Genetics l e t t e r s

The DIAGRAM-generated results for these loci indicated that three loci, d −20 −14 −11 –11 –11 –10 –8 –8 –7 –7 including the FITM2-R3HDML-HNF4A (rs6017317: P = 1.47 × 10−2,

P OR = 1.07), CMIP (rs16955379: P = 3.33 × 10−2, OR = 1.20) and MAEA (using the proxy SNP for rs6815464, rs11247991 (r2 = 0.96): 1.57 × 10 1.99 × 10 1.12 × 10 4.96 × 10 8.41 × 10 2.06 × 10 1.30 × 10 2.30 × 10 2.84 × 10 9.49 × 10 P = 6.56 × 10−3, OR = 1.19) loci, were modestly associated with T2D, whereas a locus in GLIS3 (rs7041847: P = 6.43 × 10−2, OR = 1.04) was nominally associated with T2D. The direction of effect was consistent in four (PSMD6, PEPD, WWOX and KCNK16) of the six loci that were = 1) was genotyped in the 2

r not replicated in DIAGRAM+ (Supplementary Table 5). OR (95% CI) Combined (stages 1, 2 and 3) We analyzed the functional connections among the 10 new T2D 1.13 (1.10–1.16) 1.10 (1.07–1.13) 1.09 (1.07–1.12) 1.11 (1.07–1.14) 1.09 (1.06–1.12) 1.12 (1.08–1.16) 1.10 (1.07–1.14) 1.08 (1.05–1.11) 1.08 (1.05–1.12) 1.08 (1.05–1.12)

c genes and the 28 known T2D genes that we replicated in this study −15 −9 −7 –5 –5 –6 –4 –3 –2 –2 (Supplementary Table 3) using GRAIL33. The connection results P highlighted notable biological functions for sets of genes within

4.15 × 10 2.89 × 10 3.96 × 10 2.31 × 10 1.41 × 10 3.20 × 10 5.46 × 10 3.50 × 10 2.19 × 10 1.61 × 10 T2D-associated regions (Supplementary Fig. 4 and Supplementary replication) Tables 7 and 8). For example, KCNK16 has strong connections with previously known T2D genes encoding potassium channels (KCNJ11 The proxy SNP rs9930117 ( de novo e and KCNQ1), implying that is has a physiological role in the regula- tion of potassium transport in pancreatic cells.

OR (95% CI) We examined the association between each new T2D SNP and Stage 3 ( 1.16 (1.11–1.20) 1.11 (1.07–1.15) 1.10 (1.06–1.14) 1.10 (1.05–1.15) 1.08 (1.05–1.12) 1.16 (1.09–1.23) 1.11 (1.04–1.17) 1.06 (1.02–1.10) 1.05 (1.01–1.10) 1.06 (1.01–1.11) the expression of genes within 1 Mb of these SNPs by an expres- b –3 –2 –2 –2 –1 –2 −5 −3 −2 −3 sion quantitative trait locus (eQTL) analysis using the data from the MuTHER consortium. One SNP (rs3786897) in an intron of PEPD P was highly associated with the mRNA expression of PEPD in the adi- 6.59 × 10 1.21 × 10 4.46 × 10 1.48 × 10 1.28 × 10 3.33 × 10 3.67 × 10 2.20 × 10 8.42 × 10 2.09 × 10

replication) pose tissue of 776 individuals of European ancestry (PeQTL = 2.14 × 10−8) (Supplementary Table 9). However, this SNP did not show an association with T2D in populations of European ancestry, thus the in silico importance of this finding is unclear. Up to 25,079 cases and 29,611 controls.

d We considered the possibility that autoimmune diabetes (rather OR (95% CI) Stage 2 ( than T2D) may be driving some of the signals that we observed. 1.13 (1.07–1.20) 1.09 (1.03–1.15) 1.07 (0.99–1.15) 1.11 (1.04–1.18) 1.06 (1.00–1.13) 1.09 (1.02–1.17) 1.05 (0.99–1.12) 1.07 (1.01–1.15) 1.10 (1.03–1.17) 1.09 (1.02–1.16) First, the cases from all the studies we examined predominantly had −4 −4 −5 −5 –6 –4 –6 –6 –5 –4 adult-onset diabetes (age of disease onset ≥30 years), and none of a P the clinically diagnosed subjects had T1D, which is defined by the presence of acute ketosis and the continuous requirement of insulin 8.21 × 10 1.29 × 10 2.43 × 10 6.47 × 10 4.85 × 10 1.45 × 10 3.74 × 10 5.34 × 10 2.20 × 10 1.76 × 10 beginning within 1 year after diagnosis. Second, we researched the associations for all known T1D-associated variants in our dataset. Only a small number of loci showed association after this analysis Nature America, Inc. All rights reserved. All rights Inc. America, 2 Nature

Stage 1 (discovery) (Supplementary Table 10). These results are in distinct contrast to those for known T2D-associated variants, many of which replicated OR (95% CI) © 201

Up to 12,284 cases and 13,172 controls. in our study (Supplementary Table 3), further suggesting that our 1.09 (1.04–1.14) 1.09 (1.04–1.14) 1.10 (1.05–1.15) 1.12 (1.06–1.18) 1.11 (1.06–1.17) 1.11 (1.05–1.17) 1.14 (1.08–1.20) 1.11 (1.06–1.16) 1.13 (1.07–1.20) 1.12 (1.05–1.18) c

findings are most relevant to T2D. Third, as variants close to the T T T T A C G G G G 19 allele Other GLIS3 locus have been shown to be associated with T1D , we exam- ined the association between rs7041847 and diabetes in four studies T T C A C C A C G G Risk allele (n = 8,383) in which individulas with positive glutamic acid decar- boxylase (GAD) antibodies had been excluded (as individuals with T1D are positive for GAD antibodies, whereas individuals with T2D HNF4A - are not) (data not shown). In each study, the associations between this SNP and diabetes were the same as the association found when we included all the samples (meta-analysis P = 3.4 × 10−4, OR = 1.12). R3HDML - PAX4 Nearby gene - This finding, along with the fact that SNPs at the GLIS3 locus also

Up to 5,843 cases and 4,574 controls. 3 b show associations with fasting plasma glucose in nondiabetic adults MAEA GLIS3 FITM2 GCC1 PSMD6 ZFAND3 PEPD KCNK16 CMIP WWOX and in healthy children and adolescents4, is consistent with the hypothesis that SNPs at this locus may affect fasting glucose home- ostasis rather than the immune system. Taken together, it is unlikely

2D loci reaching genome-wide significance from a combined meta-analysis of stages 1, 2 and 3 and 2 1, stages meta-analysisof combined a fromsignificancegenome-wide reaching loci 2D that a substantial proportion of the positive associations observed in Position (bp) T 1,299,901 4,277,466 42,380,380 126,952,194 64,023,337 38,214,822 38,584,848 39,392,028 80,046,874 77,964,419 our study were driven by autoimmune diabetes. 4 9 7 3 6 6 16 16

Chr. This study is the largest GWAS meta-analysis, to our knowledge, 20 19 conducted for T2D in east Asians. Findings from this study high- e ight new ight

E light not only previously unknown biological pathways but also

population-specific loci for T2D. The association of rs9470794 in ZFAND3 with T2D seems to be highly specific to east Asian able 1 able Up to 6,952 cases and 11,865 controls. T SNP Loci showing strong evidence of association with T2D rs6815464 rs7041847 rs6017317 rs6467136 rs831571 rs9470794 rs3786897 rs1535500 Loci showing moderate evidence of association with T2D rs16955379 rs17797882 a stage 3 CAGE study. populations (Supplementary Table 5), whereas the association of

Nature Genetics VOLUME 44 | NUMBER 1 | JANUARY 2012 69 l e t t e r s

a b c Recombination rate (cM/Mb)

Plotted Plotted Plotted Recombination rate (cM/Mb) Recombination rate (cM/Mb) SNPs SNPs SNPs 2 rs6815464 100 15 2 rs7041847 100 2 100 20 r r r rs6467136 0.8 0.8 10 0.8 0.6 80 0.6 80 0.6 80

P 15

0.4 0.4 8 0.4 P

0.2 60 P 10 0.2 0.2 10 60 60 10 6 10 40 10

–log 4 40 5 40 –log 20 –log 5 2 20 0 0 20

GAK FGFRL1 SPON2 C4orf42 KIAA1530 FAM53A 0 0

TMEM175 RNF212 LOC100130872 MAEA CRIPAK SLBP 0 0 GRM8 ZNF800 GCC1 SND1

DGKQ LOC100130872-SPON2 TMEM129 GLIS3 SLC1A1 C9orf68 ARF5

SLC26A1 CTBP1 TACC3 C9orf70 PPAPDC2 FSCN3

IDUA CDC37L1 PAX4

1.0 1.2 1.4 1.6 4.0 4.2 4.4 4.6 126.6 126.8 127.0 127.2 Position on chr4 (Mb) Position on chr9 (Mb) Position on chr7 (Mb) d e f Plotted Plotted Plotted Recombination rate (cM/Mb) Recombination rate (cM/Mb)

SNPs SNPs SNPs Recombination rate (cM/Mb) 12 r2 rs6017317 100 r2 rs831571 100 r2 rs9470794 100 10 10 10 0.8 0.8 0.8 0.6 80 0.6 80 0.6 80 0.4 8 0.4 8 0.4 P 8 0.2 0.2 0.2 60 60 P P 10

6 6 6 60 10 40 10 40

–log 4 4 40

–log 4 2 20 –log 2 20 20 0 0 0 0 2 TOX2 JPH2 GDAP1L1 HNF4A TTPAL PKIGWISP2 SNTN C3orf49 PSMD6 PRICKLE2 0 0 C20orf111 FITM2 SERINC3 ADA THOC7

R3HDML ATXN7 ZFAND3 BTBD9

42.0 42.2 42.4 42.6 63.8 64.0 64.2 64.4 38.0 38.2 38.4 38.6 Position on chr20 (Mb) Position on chr3 (Mb) Position on chr6 (Mb) g h i Recombination rate (cM/Mb) Plotted Plotted Recombination rate (cM/Mb) Plotted SNPs Recombination rate (cM/Mb SNPs SNPs 10 r2 rs3786897 100 10 rs1535500 r2 100 10 rs16955379 r2 100 0.8 0.8 0.8 8 0.6 80 8 0.6 80 8 0.6 80 0.4 0.4 0.4 P

P 0.2 P 0.2 0.2

6 60

6 60 6 60 10 10 10 4 40

4 40 4 40 –log –log –log 2 20 2 20 2 20 0 0

0 0 0 0 C16orf46 BCMO1 GAN CMIP PLCG2 ) RHPN2 WDR88 CEBPA CEBPG CHST8 DNAH8 C6orf64 KCNK5 KCNK17 LOC100329108

GPATCH1 LRP3 LOC80054 PEPD KCTD15 GLP1R KCNK16 GCSH

SLC7A10 KIF6 PKD1L2

38.2 38.4 38.6 38.8 39.0 39.2 39.4 39.6 79.8 80.0 80.2 80.4 Position on chr19 (Mb) Position on chr6 (Mb) j Position on chr16 (Mb) Plotted

SNPs Recombination rate (cM/Mb 2 Nature America, Inc. All rights reserved. All rights Inc. America, 2 Nature 10 rs17797882 r 100 0.8 0.6 8 0.4 80 Figure 2 Regional association plots for new T2D loci. a–j At the top, the positions of SNPs are 0.2 P

© 201 6 60

shown, and in the middle, the regional association results from the genome-wide meta-analysis 10 4 40 are shown. The trend test –log10 P values are shown for SNPs distributed in a 0.8-Mb genomic –log region centered on the most strongly associated signal, which is depicted as a purple diamond 2 20 for the stage 1 results and a red diamond for the combined stage 1, 2 and 3 results. At the bottom, the locations of known genes in the region are shown. The genetic information was 0 0 ) WWOX MAF from the build hg18, and the LD structure was based on the 1000 Genomes 77.6 77.8 78.0 78.2 Project JPT+CHB data (June 2010). Position on chr16 (Mb)

rs11634397 near ZFAND6 seems to be specific to European popula- help understand T2D phenotypes characteristic of each population, tions (Supplementary Table 3). We observed a substantial difference for example, the high rates of diabetes seen at lower average BMIs in the risk allele frequencies of both loci between the two continental in east Asians. (Asian and European) populations (rs9470794: risk allele frequency (RAF) = 0.32 for the Asian CHB/JPT HapMap population compared to URLs. IMPUTE, http://mathgen.stats.ox.ac.uk/impute/impute.html; RAF = 0.12 for the European CEU HapMap population; rs11634397: MACH, http://www.sph.umich.edu/csg/abecasis/MACH/; BEAGLE, RAF = 0.07 for CHB/JPT compared to RAF = 0.64 for CEU). Although http://faculty.washington.edu/browning/beagle/beagle.html; METAL, these loci are related to T2D differently in the two populations (the http://www.sph.umich.edu/csg/abecasis/Metal; WGAViewer, http:// ZFAND3 locus is specific to Asians, whereas the ZFAND6 locus to compute1.lsrc.duke.edu/softwares/WGAViewer/; SNAP, http://www. Europeans), these results lead to speculation that the broader A20 broadinstitute.org/mpg/snap/; LocusZoom, http://csg.sph.umich. domain-containing zinc finger protein family has a role in the etiology edu/locuszoom/; GenABEL, http://www.genabel.org/; ProbABEL, of T2D. Additional population-specific T2D loci were also suggested http://www.genabel.org/packages/ProbABEL. by our analysis, for example, WWOX (rs17797882) (Supplementary Table 5) in east Asians and ZBED3 (rs4457053) (Supplementary Methods Table 3) in Europeans. Despite the lack of clear physiological evi- Methods and any associated references are available in the online dence on T2D pathogenesis, these findings may provide clues to version of the paper at http://www.nature.com/naturegenetics/.

70 VOLUME 44 | NUMBER 1 | JANUARY 2012 Nature Genetics l e t t e r s

Note: Supplementary information is available on the Nature Genetics website. Y.L., K.H., C.H., Y.-C.C., S.H.K., A.P.M. and R.C.W.M. The data were analyzed by M.J.G., X.S., Y.J.K., R.T.H.O., W.T.T., Y.Y.T., J.L., C.-H.C., L.-C.C., Y.W., N.R.L., Y.L., Acknowledgments L.S.A., K.L.M., T.Y., C.H., Y.-C.C., S.H.K., Y.S.C., S.K., Å.K.H. and R.C.W.M. The We thank all the participants and the staff of the BioBank Japan project. The reagents, materials and analysis tools were contributed by E.S.T., B.-G.H., N.K., project was supported by a grant from the Leading Project of Ministry of D.P.-K.N., J.J.-M.L., J.L., M.S., T.A., T.Y.W., E.N., M.Y., J.N., J.J.L., W.Z., Q.C., Y.G., Education, Culture, Sports, Science and Technology Japan. W.L., F.B.H., X.O.S., F.-J.T., Y.-T.C., J.-Y.W., N.R.L., Y.L., K.O., H.I., R.T., C.W., Y.B., The Japan Cardiometabolic Genome Epidemiology (CAGE) Network Studies T.-J.C., L.-M.C., K.S.P., H.-L.K., N.H.C., J.-Y.L., W.Y.S. and J.C.N.C. The manuscript were supported by grants for the Program for Promotion of Fundamental Studies was written by Y.S.C., M.S. and E.S.T. All authors reviewed the manuscript. in Health Sciences, National Institute of Biomedical Innovation Organization (NIBIO); the Core Research for Evolutional Science and Technology (CREST) from COMPETING FINANCIAL INTERESTS the Japan Science Technology Agency; the Grant of National Center for Global The authors declare no competing financial interests. Health and Medicine (NCGM). We thank the Office of Population Studies Foundation research and data collection Published online at http://www.nature.com/naturegenetics/. teams for the Cebu Longitudinal Health and Nutrition Survey. This work was Reprints and permissions information is available online at http://www.nature.com/ supported by US National Institutes of Health grants DK078150, TW05596, reprints/index.html. HL085144 and TW008288 and pilot funds from grants RR20649, ES10126, and DK56350. 1. Kang, H.S. et al. Transcription factor Glis3, a novel critical player in the regulation We acknowledge support from the Hong Kong Government Research Grants of pancreatic beta-cell development and insulin gene expression. Mol. Cell. Biol. 29, Council Central Allocation Scheme (CUHK 1/04C), Research Grants Council 6366–6379 (2009). Earmarked Research Grant (CUHK4724/07M) and the Innovation and Technology 2. Yang, Y., Chang, B.H., Samson, S.L., Li, M.V. & Chan, L. The Kruppel-like zinc finger protein Glis3 directly and indirectly activates insulin gene transcription. Fund of the Government of the Hong Kong Special Administrative Region (ITS/ Nucleic Acids Res. 37, 2529–2538 (2009). 487/09FP). We acknowledge the Chinese University of Hong Kong Information 3. Dupuis, J. et al. New genetic loci implicated in fasting glucose homeostasis and Technology Services Center for support of computing resources. We would their impact on type 2 diabetes risk. Nat. Genet. 42, 105–116 (2010). also like to thank the dedicated medical and nursing staff at the Prince of Wales 4. Barker, A. et al. Association of genetic loci with glucose levels in childhood and Hospital Diabetes and Endocrine Centre. adolescence: a meta-analysis of over 6,000 children. Diabetes 60, 1805–1812 This work was supported by grants from Korea Centers for Disease Control and (2011). Prevention (4845-301, 4851-302, 4851-307) and an intramural grant from the 5. Takeuchi, F. et al. Confirmation of multiple risk loci and genetic impacts bya Korea National Institute of Health (2011-N73005-00), the Republic of Korea. genome-wide association study of type 2 diabetes in the Japanese population. Diabetes 58, 1690–1699 (2009). The work by the National Taiwan University Hospital was supported in part by 6. Barroso, I. et al. Population-specific risk of type 2 diabetes conferred by HNF4A the grant (NSC99-3112-B-002-019) from the National Science Council of Taiwan. P2 promoter variants: a lesson for replication studies. Diabetes 57, 3161–3165 We would also like to acknowledge the National Genotyping Center of National (2008). Research Program for Genomic Medicine (NSC98-3112-B-001-037), Taiwan. 7. Silander, K. et al. Genetic variation near the hepatocyte nuclear factor-4 α gene The work by the Shanghai Diabetes Genetic Study was supported in part by predicts susceptibility to type 2 diabetes. Diabetes 53, 1141–1149 (2004). the US National Institutes of Health grants R01CA124558, R01CA64277, 8. Zimmet, P., Alberti, K.G. & Shaw, J. Global and societal implications of the diabetes R01CA70867, R01CA90899, R01CA100374, R01CA118229, R01CA92585, UL1 epidemic. Nature 414, 782–787 (2001). RR024975, DK58845 and HG004399, the Department of Defense Idea Award 9. Tkác, I. Metabolic syndrome in relationship to type 2 diabetes and . Diabetes Res. Clin. Pract. 68 (suppl. 1), S2–S9 (2005). BC050791, Vanderbilt Ingram professorship funds and the Allen Foundation 10. Prokopenko, I., McCarthy, M.I. & Lindgren, C.M. Type 2 diabetes: new genes, new Fund. We thank the dedicated investigators and staff members from research understanding. Trends Genet. 24, 613–621 (2008). teams at Vanderbilt University, Shanghai Cancer Institute and the Shanghai 11. Rung, J. et al. Genetic variant near IRS1 is associated with type 2 diabetes, insulin Institute of Preventive Medicine, and especially, the study participants for their resistance and hyperinsulinemia. Nat. Genet. 41, 1110–1115 (2009). contributions in the studies. 12. Manolio, T.A. et al. Finding the missing heritability of complex diseases. The work of the Shanghai Diabetes Study was supported by grants from the Nature 461, 747–753 (2009). National 973 Program (2011CB504001), the Project of National Natural Science 13. Yasuda, K. et al. Variants in KCNQ1 are associated with susceptibility to type 2 Foundation of China (30800617) and the Shanghai Rising-Star Program diabetes mellitus. Nat. Genet. 40, 1092–1097 (2008). Nature America, Inc. All rights reserved. All rights Inc. America, 2 Nature 14. Unoki, H. et al. SNPs in KCNQ1 are associated with susceptibility to type 2 diabetes (09QA1404400), China. in east Asian and European populations. Nat. Genet. 40, 1098–1102 (2008). The work of the Shanghai Jiao Tong University Diabetes Study was supported by 15. Yamauchi, T. et al. A genome-wide association study in the Japanese population

© 201 grants from the National 863 Program (2006AA02A409) and the major program of identifies susceptibility loci for type 2 diabetes at UBE2E2 and C2CD4A-C2CD4B. the Shanghai Municipality for Basic Research (08dj1400601), China. Nat. Genet. 42, 864–868 (2010). The work of the Seoul National University Hospital was supported by grants from 16. Tsai, F.J. et al. A genome-wide association study identifies susceptibility variants the Korea Health 21 R&D Project, Ministry of Health & Welfare (00-PJ3-PG6- for type 2 diabetes in Han Chinese. PLoS Genet. 6, e1000847 (2010). GN07-001) and the World Class University project of the Ministry of Education, 17. Shu, X.O. et al. Identification of new genetic risk variants for type 2 diabetes. PLoS Science and Technology (MEST) and National Research Foundation (NRF) Genet. 6, e1001127 (2010). 18. Stommel, M. & Schoenborn, C.A. Variations in BMI and prevalence of health risks (R31-2008-000-10103-0), Korea. The Singapore Prospective Study Program (SP2) in diverse racial and ethnic populations. Obesity (Silver Spring) 18, 1821–1826 was funded through grants from the Biomedical Research Council of Singapore (2010). (BMRC05/1/36/19/413 and 03/1/27/18/216) and the National Medical Research 19. Barrett, J.C. et al. Genome-wide association study and meta-analysis find that over Council of Singapore (NMRC/1174/2008). E.S.T. also receives additional support 40 loci affect risk of type 1 diabetes. Nat. Genet. 41, 703–707 (2009). from the National Medical Research Council through a clinicians scientist award 20. Kadereit, B. et al. Evolutionarily conserved gene family important for fat storage. (NMRC/CSA/008/2009). The Singapore Malay Eye Study (SiMES) was funded Proc. Natl. Acad. Sci. USA 105, 94–99 (2008). by the National Medical Research Council (NMRC0796/2003 and NMRC/ 21. Nakajima, H. et al. Hepatocyte nuclear factor-4 α gene mutations in Japanese STaR/0003/2008) and Biomedical Research Council (BMRC, 09/1/35/19/616). non-insulin dependent diabetes mellitus (NIDDM) patients. Res. Commun. Mol. Pathol. Pharmacol. 94, 327–330 (1996). Y.Y.T. acknowledges support from the Singapore National Research Foundation, 22. Johansson, S. et al. Studies in 3,523 Norwegians and meta-analysis in 11,571 NRF-RF-2010-05. The Genome Institute of Singapore carried out all the subjects indicate that variants in the hepatocyte nuclear factor 4 α (HNF4A) P2 genotyping for the samples from Singapore also provided funding for the region are associated with type 2 diabetes in Scandinavians. Diabetes 56, genotyping of the samples from SP2. 3112–3117 (2007). 23. Girard, C. et al. Genomic and functional characteristics of novel human pancreatic AUTHOR CONTRIBUTIONS 2P domain K+ channels. Biochem. Biophys. Res. Commun. 282, 249–256 The study was supervised by E.S.T., B.-G.H., N.K., Y.S.C., Y.Y.T., W.Z., Q.C., (2001). X.O.S., Y.-T.C., J.-Y.W., L.S.A., K.L.M., T.K., C.H., W.J., L.-M.C., Y.M.C., K.S.P., 24. Ashcroft, F.M. ATP-sensitive potassium channelopathies: focus on insulin secretion. J.-Y.L. and J.C.N.C. The experiments were conceived of and designed by Y.S.C., J. Clin. Invest. 115, 2047–2058 (2005). 25. Soni, S. et al. Absence of erythroblast macrophage protein (Emp) leads to failure E.S.T., N.K., D.P.-K.N., J.J.-M.L., M.S., T.Y.W., Y.Y.T., W.Z., F.B.H., X.O.S., C.-H.C., of erythroblast nuclear extrusion. J. Biol. Chem. 281, 20181–20189 (2006). F.-J.T., Y.-T.C., J.-Y.W., L.S.A., K.L.M., S.M., C.H., L.-M.C., K.S.P., M.J.G., M.I.M. 26. Nam, D., Kim, J., Kim, S.Y. & Kim, S. GSA-SNP: a general approach for gene set and R.C.W.M. The experiments were performed by J.L., M.S., J.J.L., J.-Y.W., S.M., analysis of polymorphisms. Nucleic Acids Res. 38, W749–W754 (2010). R.Z., K.Y., Y.-C.C., T.-J.C., L.-M.C. and S.H.K. Statistical analyses was performed 27. Scott, L.J. et al. A genome-wide association study of type 2 diabetes in Finns by M.J.G., X.S., Y.J.K., R.T.H.O., W.T.T., Y.Y.T., F.T., J.L., C.-H.C., L.-C.C., Y.W., detects multiple susceptibility variants. Science 316, 1341–1345 (2007).

Nature Genetics VOLUME 44 | NUMBER 1 | JANUARY 2012 71 l e t t e r s

28. Luke, M.R., Houghton, F., Perugini, M.A. & Gleeson, P.A. The trans-Golgi network 31. Voight, B.F. et al. Twelve type 2 diabetes susceptibility loci identified through large- GRIP-domain form α-helical homodimers. Biochem. J. 388, 835–841 scale association analysis. Nat. Genet. 42, 579–589 (2010). (2005). 32. Frayling, T.M. et al. A common variant in the FTO gene is associated with body 29. Jo, W., Endo, M., Ishizu, K., Nakamura, A. & Tajima, T. A novel PAX4 mutation mass index and predisposes to childhood and adult obesity. Science 316, 889–894 in a Japanese patient with maturity-onset diabetes of the young. Tohoku J. Exp. (2007). Med. 223, 113–118 (2011). 33. Raychaudhuri, S. et al. Identifying relationships among genomic disease regions: 30. Wang, X. et al. Mass spectrometric characterization of the affinity-purified human predicting genes at pathogenic SNP associations and rare deletions. PLoS Genet. 5, 26S proteasome complex. Biochemistry 46, 3553–3565 (2007). e1000534 (2009).

1Center for Genome Science, National Institute of Health, Osong Health Technology Administration Complex, Chungcheongbuk-do, Cheongwon-gun, Gangoe-myeon, Yeonje-ri, Korea. 2Institute of Biomedical Sciences, Academia Sinica, Nankang, Taipei, Taiwan. 3School of Chinese Medicine, China Medical University, Taichung, Taiwan. 4Shanghai Diabetes Institute, Shanghai Key Laboratory of Diabetes Mellitus, Department of Endocrinology and Metabolism, Shanghai Jiao Tong University Affiliated Sixth People’s Hospital, Shanghai, China. 5Department of Medicine, Vanderbilt Epidemiology Center, Vanderbilt-Ingram Cancer Center, Vanderbilt University School of Medicine, Nashville, Tennessee, USA. 6Graduate School for Integrative Science and Engineering, National University of Singapore, Singapore, Singapore. 7Centre for Molecular Epidemiology, National University of Singapore, Singapore, Singapore. 8Research Institute, National Center for Global Health and Medicine, Shinjuku-ku, Tokyo, Japan. 9Department of Genetics, University of North Carolina, Chapel Hill, North Carolina, USA. 10Department of Diabetes and Metabolic Diseases, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan. 11Department of Internal Medicine, National Taiwan University Hospital, Taipei, Taiwan. 12Department of Internal Medicine, Seoul National University College of Medicine, Seoul, Korea. 13Department of Medicine and Therapeutics, Chinese University of Hong Kong, Prince of Wales Hospital, Hong Kong, China. 14Division of Genome Analysis, Research Center for Genetic Information, Medical Institute of Bioregulation, Kyushu University, Higashi-ku, Fukuoka, Japan. 15Department of Nutrition, University of North Carolina, Chapel Hill, North Carolina, USA. 16Singapore Eye Research Institute, Singapore National Eye Centre, Singapore, Singapore. 17Department of Ophthalmology, National University of Singapore, Singapore, Singapore. 18Department of Epidemiology, Shanghai Cancer Institute, Shanghai, China. 19Department of Nutrition and Epidemiology, Harvard School of Public Health, Boston, Massachusetts, USA. 20Department of Biochemistry, School of Medicine, Ewha Womans University, Seoul, Korea. 21School of Systems Biomedical Science, Soongsil University, Dongjak-gu, Seoul, Korea. 22Department of Epidemiology and Public Health, National University of Singapore, Singapore, Singapore. 23Office of Population Studies Foundation Inc., University of San Carlos, Cebu City, Philippines. 24Department of Biostatistics, University of North Carolina, Chapel Hill, North Carolina, USA. 25Genome Institute of Singapore, Agency for Science, Technology and Research, Singapore, Singapore. 26Shanghai Institute of Preventive Medicine, Shanghai, China. 27Division of Endocrinology and Diabetes, Department of Internal Medicine, Nagoya University Graduate School of Medicine, Nagoya, Japan. 28Department of Diabetes and Endocrinology, Chubu Rosai Hospital, Nagoya, Japan. 29Centre for Eye Research Australia, University of Melbourne, East Melbourne, Victoria, Australia. 30Department of Genome Science, Aichi-Gakuin University, School of Dentistry, Nagoya, Japan. 31Department of Geriatric Medicine, Graduate School of Medical Sciences, Kyushu University, Higashi-ku, Fukuoka, Japan. 32Department of Endocrinology, Metabolism and Diabetes, Kinki University School of Medicine, Osaka-sayama, Osaka, Japan. 33Department of Preventive Medicine, Ajou University School of Medicine, Suwon, Korea. 34Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford, UK. 35Oxford Centre for Diabetes, Endocrinology and Metabolism, University of Oxford, Churchill Hospital, Oxford, UK. 36A full list of members is provided in the Supplementary Note. 37Department of Internal Medicine and Bioregulatory Science, Graduate School of Medical Sciences, Kyushu University, Higashi-ku, Fukuoka, Japan. 38World Class University program, Department of Molecular Medicine and Biopharmaceutical Sciences, Graduate School of Convergence Science and Technology and College of Medicine, Seoul National University, Seoul, Korea. 39Graduate Institute of Clinical Medicine, National Taiwan University School of Medicine, Taipei, Taiwan. 40Laboratory for Endocrinology and Metabolism, RIKEN Center for Genomic Medicine, Yokohama, Japan. 41Department of Statistics and Applied Probability, National University of Singapore, Singapore, Singapore. 42Department of Medicine, National University of Singapore, Singapore, Singapore. 43Duke-National University of Singapore Graduate Medical School, Singapore, Singapore. 44Institute for Human Genetics, University of California, San Francisco, California, USA. 45Blood Systems Research Institute, San Francisco, California, USA. 46These authors contributed equally to this work. 47These authors jointly directed this work. Correspondence should be addressed to Y.S.C. ([email protected]), M.S. ([email protected]) or E.S.T. ([email protected]). Nature America, Inc. All rights reserved. All rights Inc. America, 2 Nature © 201

72 VOLUME 44 | NUMBER 1 | JANUARY 2012 Nature Genetics ONLINE METHODS not correct for genomic control in the stage 1 analyses because the infla- Study subjects. Stage 1 subjects were drawn from eight T2D GWAS par- tion was modest, suggesting that population structure is unlikely to cause ticipating in the AGEN consortium, which was organized to enable genetic substantial inflation of the stage 1 results (Supplementary Table 2). The studies on diverse complex traits in 2010. These eight studies included 6,952 selection criteria for the lead SNPs to take forward to stage 2 in silico repli- cases with T2D and 11,865 controls from the Korea Association Resource cation analysis were as follows: (i) stage 1 meta-analysis P < 5 × 10−4 (based Study (KARE), the Singapore Diabetes Cohort Study (SDCS), the Singapore on the divergence between the observed and expected P values on the Prospective Study Program (SP2), the Singapore Malay Eye Study (SiMES), quantile-quantile plot; Supplementary Fig. 2); (ii) heterogeneity P > 0.01; the Japan Cardiometabolic Genome Epidemiology Network (CAGE), the and (iii) at least seven studies having been included in the stage 1 meta- Shanghai Diabetes Genetic Study (SDGS), the Taiwan T2D Study (TDS) and analysis (Supplementary Table 4). After removing known variants associated the Cebu Longitudinal Health and Nutritional Survey (CLHNS). Subjects in with T2D, proxies for each lead SNP (r2 > 0.8) were selected using the SNAP stage 2 included 5,843 cases with T2D and 4,574 controls from three inde- software (see URLs). The replication genotyping for stage 3 was performed pendent GWAS, the BioBank Japan Study (BBJ), the Health2 T2D Study for the new SNPs having a stage 2 combined P < 10−5. Regional association (H2T2DS) and the Shanghai Jiao Tong University Diabetes Study (SJTUDS), results from genome-wide meta-analysis were plotted using LocusZoom for in silico replication analysis. Stage 3 included up to 12,284 cases with T2D software (see URLs) for SNPs reaching genome-wide significance from the and 13,172 controls from five different studies, the Japan Cardiometabolic combined meta-analysis of stages 1, 2 and 3. Genome Epidemiology Network (CAGE), the Shanghai Diabetes Study I/II (SDS I/II), the Chinese University of Hong Kong Diabetes Study (CUHKDS), Principal components analysis. A list of 76,534 common SNPs across the the National Taiwan University Hospital Diabetes Study (NTUHDS) and the Illumina 550, 610 and 1M and Affymetrix 5.0 and 6.0 arrays were first selected. Seoul National University Hospital Diabetes Study (SNUHDS), for de novo This set of SNPs in the Asian (CHB+JPT) HapMap II samples was then trained replication analysis. The study design and T2D diagnosis criteria used in each to generate a list of 44,524 SNPs having pairwise LD < 0.3 in a sliding window study included in stages 1, 2, and 3 are described in Supplementary Table 1 of 50 SNPs. Individuals from each component study and from HapMap II and the Supplementary Note. Each study obtained approval from the appro- were plotted based on the first two eigenvectors produced by the principal priate institutional review boards of each participating institution, and written components analysis. informed consent was obtained from all participants. The three-stage design of the overall study is depicted in Supplementary Figure 1. eQTL analysis. Gene expression information from 776 adipose tissues, 667 skin tissues and 777 lymphoblastoid cell lines was obtained from the MuTHER Genotyping and imputation. Subjects for the stage 1 and 2 analyses were consortium36. The eQTL data for eight of the ten T2D loci identified in this genotyped with high-density SNP typing platforms that covered the entire study were available in the MuTHER dataset. Most of those loci passed the human genome. In most of the studies, only unrelated samples with missing filtering criteria, such as MAF > 5% and INFO > 0.8, except for rs16955379, genotype call rates below 5% were included for subsequent GWAS analyses. which has MAF = 1.5% in the MuTHER data set. Two of the ten loci that were For the genome-wide association meta-analysis, each study participating in used in the eQTL analysis, rs6815464 (on chromosome 4) and rs17797882 stages 1 and 2 performed SNP imputation. IMPUTE, MACH or BEAGLE (see (on chromosome 16), are not included in the MuTHER data set. Association URLs) were used, together with haplotype reference panels from the JPT and between each SNP with a significant association to T2D and the normal- CHB samples that are available in the HapMap database (JPT+CHB+CEU ized mRNA expression values of genes within 1 Mb of the lead SNP were and/or YRI, in some studies) on the basis of HapMap build 36 (release 21, performed with the GenABEL and ProbABEL package (see URLs) using the 22, 23a or 24). Only imputed SNPs with high genotype information content polygenic linear model incorporating a kinship matrix in GenABEL followed (proper info > 0.5 for IMPUTE and Rsq > 0.3 for MACH and BEAGLE) were by the ProbABEL mmscore test with imputed genotypes. A multiple-testing used for the association analysis. Genotyping for the stage 3 analysis was car- correction was applied to the cis association results. P value thresholds of ried out using TaqMan, Sequenom MassARRAY or the Beckman SNP Stream P = 5.06 × 10−5 in , P = 3.81 × 10−5 in skin and P = 7.80 × 10−5

Nature America, Inc. All rights reserved. All rights Inc. America, 2 Nature method. All SNPs included in stage 3 had a genotype success rate of >98% in lymphoblastoid cell lines correspond to an estimated genome-wide false (Supplementary Table 2). discovery rate of 1%. © 201 Statistical analyses, analysis tools and SNP prioritization for stages 2 and 3. Gene relationships among implicated loci (GRAIL) analysis. A GRAIL Associations between SNPs and T2D were tested by logistic regression with analysis was performed as described previously31,33. A total of 38 genes within an additive model (1 degree of freedom) after adjustment for sex. Other T2D-associated regions were selected for the analysis. Among these genes, adjustments were permitted according to the situations in the individual 28 were from the previously implicated set (Supplementary Table 3), and studies. The meta-analysis was performed using an inverse-variance method the other 10 genes were newly implicated in this study (Table 1). PubMed assuming fixed effects, with a Cochran’s Q test to assess between-study abstracts published after December 2006 were omitted from the analysis to heterogeneity. METAL software (see URLs) was used for all meta-analyses. reduce confounding by results from T2D GWAS. A plot of the negative log of the association results from the stage 1 meta- analysis, by chromosome, was generated using WGAViewer software (see URLs).The quantile-quantile plot was constructed by plotting the distribu- 34. Hyndman, R.J. & Fan, Y. Sample quantiles in statistical packages. Am. Stat. 50, tion of observed P values for the given SNPs against the theoretical distri- 361–365 (1996). 34 35. Devlin, B., Roeder, K. & Wasserman, L. Genomic control, a new approach to genetic- bution of the expected P values for T2D . The genomic control inflation based association studies. Theor. Popul. Biol. 60, 155–166 (2001). 2 factor, λ, was calculated by dividing the median χ statistics by 0.456 (ref. 35) 36. Nica, A.C. et al. The architecture of gene regulatory variation across multiple human for individual GWAS, as well as for the stage 1 GWAS meta-analysis. We did tissues: the MuTHER study. PLoS Genet. 7, e1002003 (2011).

doi:10.1038/ng.1019 Nature Genetics